Nonparametric (NP) approaches estimate evapotranspiration (ET) based on the surface energy balance and conventional equilibrium evaporation, eliminating the need for empirical parameterization of aerodynamic or surface resistances. However, conventional NP methods often systematically underestimate latent heat flux (LE) under high evaporative demand, particularly in arid and semi-arid areas. This bias arises because conventional approaches implicitly rely on a linearized Clausius-Clapeyron relationship to represent atmospheric moisture conditions. In this study, we developed a modified nonparametric (MNP) method by restoring the nonlinear Clausius-Clapeyron relationship. The performance of the MNP method was evaluated using ground observations at three flux sites in the Heihe River Basin. The results show that the MNP method improves daily LE estimation compared to the original NP method, reducing the Root Mean Square Error (RMSE) by 19.33% (from 32.87 W & sdot;m-2 to 26.53 W & sdot;m-2) and improving the consistency of LE estimation across different sites. Under high evaporative demand conditions, where discrepancies between theoretical and actual vapour pressure deficit are large, the RMSE was reduced by 24.84%, 26.64%, and 41.58% across the Daman, Arou, and Yakou sites. Sensitivity analysis shows that the LE estimates from the MNP method are most sensitive to net radiation and temperature inputs. Overall, restoring the nonlinear Clausius-Clapeyron relationship enables a more realistic representation of atmospheric moisture constraints and significantly enhances the robustness of nonparametric LE estimation.